The Effect of Competition on Terrorist Group Operations

Stephen NemethDepartment of Political Science

244 Waters HallKansas State UniversityManhattan, KS 66506

snemeth@k-state.edu

Abstract: Scholars have long accepted the contention that competition amongst terrorist organizations raises the level of violence used by the competitors (Crenshaw, 1981, 1985, Oots, 1989; Bloom, 2004, 2005; Chenoweth, 2010). This paper discusses this claim and advances another – that competition amongst terrorist organizations creates incentives to use less violence. Using insights from the organizational ecology literature - namely that competition occurs within “species” – I create a variable that assesses intra-species competition. I test both claims using a dataset of domestic terrorism created from the Global Terrorism Dataset for the years 1970-1997. I find support for the hypothesis that competition leads to more terrorism, validating the claims of outbidding theorists. Furthermore, ideologies have differential effects on whether outbidding occurs, with nationalist and religious terrorist groups responding to competition with more terrorism and left-wing organizations responding with less.

The idea that competition amongst terrorist organizations results in a number of

negative outcomes, such as suicide terrorism, civilian targeting, and extreme violence has

become a powerful and persuasive idea in the terrorism literature (Crenshaw, 1981, 1985;

Oots, 1989; Bloom, 2005; Chenoweth, 2010). Organizations, vying for the attention of

recruits or for the support of wavering adherents, use escalating levels of violence, or

“outbidding” as a means to demonstrate their commitment and capability (Bloom, 2005).

This violence can also be turned against competing organizations, rendering the states

most likely to be fertile grounds for terrorism the most dangerous for terrorist

organizations (Oots, 1989). The outcome of this process is likely to have an indelible

mark on affected states; competition acts to legitimize more violence, encourage violence

against civilians and, in some cases, to encourage the use of suicide terror (Bloom, 2005).

Intuitively, this perspective has great appeal. Organizations that can demonstrate

their ability through continued and notable violence are advantaged in both recruitment

and group maintenance (Crenshaw, 1985; Post et al., 2003). Recruits are drawn to the

organizations which have the most selective incentives to offer, with the opportunity to

participate in violent activity forming a primary motivation (Crenshaw, 1985). Group

cohesion and retention is increased because violence and the number of operations

performed help “build morale within the membership through the experience of

cooperative operations” (Waugh, 1983: 9). Violence is currency in this perspective –

groups that are the most adept in its use ensure their survival, develop into capable

adversaries of the state, and become increasingly able to achieve their political goals.

At the same time, the argument is one of nuance. For Bloom (2005), the type of

violence characterized by outbidding only flourishes and resonates in particular

1

situations. Organizations that engage in suicide terrorism, or excessive levels of

violence, in the incorrect environment fail to win public support and potentially endanger

the organization. The terrorism literature is replete with accounts of groups

miscalculating the amount of violence a populace was willing to accept (see Ross and

Gurr, 1989; Cronin, 2006). In such cases, violence has played a negative role in the

development of the organization, leading to less recruiting success, reduced public

support, and the decline, if not the dissolution, of the group.

This paper seeks to address why competition leads to more terrorism in some

cases and less in others. Like Bloom (2005), I forward that two factors - government

policy and social acceptability – determine whether outbidding actually occurs. In areas

where those factors are present competition leads to outbidding. Areas with competition

that are absent those factors will not demonstrate outbidding behavior. In fact,

competition in such states may actually lead to less terrorism. This insight distinguishes

Bloom’s (2005) discussion of outbidding from earlier ones (Crenshaw, 1981, 1985; Oots,

1989). I test this contention using cross-national domestic terrorism data from the Global

Terrorism Dataset (GTD) for the years 1970-1997.

Second, I test these claims using an improved operational definition of

competition. Building upon the organizational ecology literature, I develop a measure of

competition that is “species” specific (Lowery and Gray, 1995). This means that

competition is restricted to groups of a similar type; in this case, competition is restricted

to groups that share a similar ideology. This lets us avoid treating two very different

concepts the same way: the absolute number of terrorist organizations and the density of

2

the organizational environment. By constructing an indicator using the latter, we achieve

a closer association between measure and reality (Post et al., 2003)

The paper is composed as follows: In the next section, I discuss the various ways

that competition may affect terrorist organizations. I highlight Bloom’s (2005) theory of

outbidding and the factors that make it possible. I follow with a discussion of the

relationship between competition and terrorism when those factors are not present and

how, in those circumstances, competition may reduce terrorism. The section ends with

the discussion of the hypotheses that stem from the theory. In the third section, I discuss

the data and coding rules. I emphasize the organizational ecology literature to create a

measure of terrorist organizational competition independent of the total number of

terrorist organizations that reside within a state. The fourth section presents empirical

tests of the hypothesis. The fifth section concludes.

THE FACETS OF COMPETITION

While outbidding has become the dominant perspective relating competition to

terrorism, the possibility exists that competition can also have a negative, rather than a

positive effect. One of the most prominent of these alternatives is that of backlash –

groups select a target or engage in violence that reduces the support of previously

sympathetic population, thus reducing the flow of resources and the ability of the

organization to carry out attacks (Ross and Gurr, 1989; Cronin, 2006). At the same time,

competition may also lead to a reduction in violence if it leads to cooperation between

two terrorist organizations or if it acts to reduce the capacity of the competitors. These

alternatives stand in contrast to the outbidding perspective, yet neither these alternatives

3

nor the outbidding perspective have been subject to quantitative analyses (although see

Chenoweth, 2010). Before these perspectives are tested, I discuss them in detail below.

Outbidding

The outbidding perspective arises from the realization that dual pressures exist

within all terrorist organizations. Terrorist leaders, “must cope with a constant tension

between their desires to preserve the organization and the membership’s desire for

action” (Crenshaw, 1985: 476). Early theories of outbidding posited that competition

solely affected these two factors (Crenshaw, 1985; Oots, 1989). Later theories, notably

that of Bloom (2004, 2005), sees competition leading to outbidding when external

conditions – the acceptability of violence and government policy - are favorable.

The first theory of outbidding sees competition contributing to the collective

radicalization of the terrorist organization (Crenshaw, 1985). Groups facing competition

from a more extreme competitor typically radicalize and engage in more violence to

prevent the defection of members to the competitor organization.i The adoption of

terrorism by the Official IRA in response to the same decision by the Provisional IRA is

indicative of this type of outbidding logic (Crenshaw, 1985; Moloney, 2010). In the

Palestinian areas, the adoption of suicide terror by secular groups after its use by

fundamentalist organizations may be another example.ii

Secondly, Crenshaw (1985) suggests that outbidding may occur when

competition, and the threat of member exit, leads members to fight harder to validate

their investment in joining the organization. The decision to join a terrorist group entails

a sacrifice and a distinct change in lifestyle; Wolf (1978: 176) depicts the life of a

terrorist as one “characterized by the lack of comfort, the absence of expendable income,

4

and the denial of leisure activity and personal privacy”. As a result, the high personal

cost endured by members makes activity their only sustenance and identity. Increased

action therefore, is a logical result of individuals seeking to ensure their choices to

engage in terrorism are justified (Kellen, 1979; Crenshaw, 1985).

The failure of the Israeli-Palestinian peace process in the late 1990s and the

subsequent rise in suicide terrorism provided the main impetus for the most recent

incarnation of the outbidding theory (Bloom, 2005). The increase of the use of suicide

terror, and the support groups received for their actions soon created an environment

where all groups had to engage in spectacular acts of violence.iii The logic of this theory

is simple and persuasive - “if multiple insurgent groups are competing for public support,

bombings will intensify in scope and number as they become both the litmus test of

militancy and the way to mobilize greater numbers of people within their community”

(Bloom, 2005: 78).iv This is known to be effective; Jerrold Post and his research team

found that nearly 60 percent of members in secular groups and 43 percent of religious

group members admitted to joining the most active terrorist organization in their

community (2003: 173).

This theory does differ from its predecessors. For Bloom (2005), outbidding

requires that the environment be able to support a group’s decision to engage in extreme

violence. In the case of the Israeli-Palestinian crisis, outbidding was successful because

the ongoing conflict, the collapse of the peace process, and the ineffectiveness of the

Palestinian Authority created an atmosphere of hopelessness that both spurred the

creation of terrorist organizations and legitimated terrorism, particularly suicide

terrorism. The newfound acceptability of this tactic became the mark by which all

5

organizations were judged, setting off the cycle of outbidding. Support for suicide

terrorism increased quickly after these occurrences, reaching upwards of 85 percent in

October 2001 from lows of 24 and 33 percent during the period from 1997 to 1999

(Bloom, 2005: 193).

In Sri Lanka, another of Bloom’s (2005) case studies, outbidding arose for largely

the same reasons. The Tamil history with the Sri Lankan government had been marked

by political futility and communal violence at the hands of the majority Sinhalese. As a

result, Tamil quality of life had changed little since Sri Lankan independence. This status

quo led to the acceptance of terrorism as a political tactic and, moreover, to the creation

of a number of Tamil separatist groups, one of which was the Liberation Tigers of Tamil

Eelam (LTTE). As a result, the violence that resulted was similar in origin to the

Palestinian territories – groups competed with one another for support within an

atmosphere that accepted this kind of violence.

Unlike the Palestinian case, the acceptability of terrorism began to wane in Sri

Lanka. The LTTE had used violence to great effect; it struck at the Sinhalese, moderate

Tamils, and its competitors to become the predominant organization pressing for Tamil

separatism. By 2001, the long-running conflict between it and the Sri Lankan

government has sapped the will of the organization’s supporters, driving it to negotiations

with the government. By the time of Bloom’s analysis, Tamil hopefulness for peace and

the end of competition between the various Tamil organizations had largely ruled out the

use of suicide terrorism as a tactic.

This perspective has led to some quantitative work – most notably that of

Chenoweth (2010). In this work, she argues that intergroup competition explains why

6

democracies experience more terrorist activity than non-democracies. She finds support

that transnational terrorist incidents and domestic terrorist organizations are more likely

to originate in competitive polities. While her analysis does provide some validation of

Bloom (2005), it is not a test of terrorist competition per se. Instead, competition

between terrorist organizations is subsumed into a larger measure of political

competition. States may, in fact, have high levels of terrorist competition while having

little political competition. Competition is also assumed to occur across the terrorist

group system. The assumption here is that resources and recruits are indifferent about

which organization they are directed to as long as they are directed to the most active

one. I discuss the potential drawbacks to this approach, particularly in the context of the

organizational ecology literature, in the methods section.

Competition and Moderation

While Bloom’s (2005) theory provides an excellent explanation of terrorist

competition and its outcomes in those areas which include both components, it is

important to consider what the effect of competition may be when either of those two

conditions is absent. First, competition may raise the cost of action; civilians see

outbidding as escalating savagery rather than meaningful attempts for political change,

driving civilians away and increasing the risk of backlash (Ross and Gurr, 1989). Second

and relatedly, this cycle of violence may serve as an implicit government counterterror

policy. Internecine struggles amongst terrorist organizations may drain groups of

resources and recruits, decreasing the efforts needed by incumbent governments for

counterterrorism. These considerations may lead to the observation that competition may

lead to less, rather than more, violence.

7

The possibility of backlash is a major risk of outbidding in an incorrect

environment. Groups have frequently miscalculated the effects of their violence and

triggered this response. Cronin (2006) points out the Popular Front for the Liberation of

Palestine – General Command (PFLP-GC) and ETA as two groups amongst many that

have engendered condemnation for their actions. In the context of outbidding, the

Omagh bombing in August 1998 can be seen as a major contemporary instance of

backlash occurring from competition –between the Real Irish Republican Army, the

perpetrators, and the Provisional Irish Republican Army (Dingley, 2001). The deaths of

29 civilians from the attack, precisely during the Northern Ireland peace process, led to a

denunciation of the perpetrators from every political party in Northern Ireland, Ireland,

and Great Britain. Moreover, group support from individuals within the United States

waned and support for the then-recently signed Good Friday agreement increased

(Dingley, 2001).

Lastly, competition and outbidding may be damaging because it works to

inadvertently advance the state’s agenda rather than the terrorists’ goals. In Sri Lanka

and Turkey, the government either funded a favored terrorist organization or turned a

blind eye to intergroup violence in the hopes that the organizations would eliminate, or

greatly weaken, each other (Bloom, 2005). The extent to which groups may be

concerned with this problem is unknown. However, the decision of the PKK and Turkish

Hezbollah to abandon intergroup conflict in favor of jointly fighting the government may

indicate that groups have some realization of the dangers of conflict with one another

(Bloom, 2005).

8

These concerns all contribute to a number of mechanisms which may lead to

competition causing less violence. Groups may first simply try to avoid backlash by

changing the nature and number of their attacks. This is especially evident if we consider

the level of public support as relatively fixed; each increase of an additional

organization reduces each group’s potential allocation of support.

Given the assumption that using outbidding violence results in a net

gain of organizational support for the organization, the use of this

strategy in competitive situations ultimately results in each group

gaining a pool of supporters that does not compensate for the loss of

members. Instead, this should lead to the formation of coalitions - a

rare occurrence (Crenshaw, 1985; Oots, 1989) - or the moderation of

violence.

An additional explanation for why competition may lead to less violence may lie

in the work of Lewis Coser (1956). He argued that external conflict can result in

cohesion within groups or the creation of coalitions, or at least elements of cooperation,

amongst different groups. In terms of this analysis, the increased attention brought to

groups by competition can lead to elements of coordination, or even cooperation,

between the competitors. Bloom (2005) noted that both the Kurdish Workers’ Party

(PKK) and Turkish Hezbollah, when confronted with government attention, abandoned

outbidding to adopt a campaign of violence directed against the Turkish government.

Brym and Araj (2008: 493-495) note that Israeli action had even pushed both Hamas and

Fatah into situations of tactical cooperation. In both cases, the number of terrorist events

perpetrated by these groups was lower after coordination than they were prior.v

9

The two perspectives regarding terrorist group competition provide contrasting

hypotheses about their effect on group operations. They are presented as follows:

Hypothesis 1: Groups in competitive and favorable environments will commit more terrorist acts than groups in non-competitive and non-favorable environments.

Hypothesis 2: Groups in competitive and favorable environments will commit less terrorist acts than groups in non-competitive and non-favorable environments.

DATA AND METHODS

In order to analyze the relationships discussed here, I use the Global Terrorism

Dataset (GTD) from the National Consortium for the Study of Terrorism and Responses

to Terrorism (START) at the University of Maryland. The dataset includes 81,800 acts

of domestic and international terrorism occurring in 180 states spanning the years 1970-

2007.vi

The GTD provides a remarkably flexible platform for two reasons. The first is

that its definition of terrorism is quite broad, referring to terrorism as “the threatened or

actual use of illegal force and violence to attain a political, economic, religious, or social

goal through fear, coercion, or intimidation” (LaFree and Dugan, 2007: ii). The

definition, absent references to targets and referring to a variety of goals and

mechanisms, allows researchers to create their own definitions and circumvents the

contentious issue of having to use an established definition of terrorism (Schmid and

Jongman, 1988; Hoffman, 1998; Silke 2004). Second, the GTD contains a wide variety

of perpetrators of terrorist violence. Well-defined organizations such as Hamas, Al-

Qaeda, and the Tupamaros are included along with actors such as political parties,

student protesters, and rebels. This also provides flexibility by allowing researchers to

focus on interests ranging from crime to social movements.

10

Since the analyses regard the effects of competition on group behavior, I use the

group-state/year as the unit of analysis. This distinction, rather than group-year, accounts

for the behavior of the same group in different states, such as Karen National Union in

Myanmar versus its branch in Thailand. The repressive environment in Myanmar is

likely to exert a different effect on group behavior than the somewhat more permissive

environment in Thailand.

Second, this also accounts for differing levels of competition within the same

state. A state may have a competitive atmosphere regarding left-wing groups, but a

relatively monopolistic environment regarding groups on the right. Accounting for

competition as a state-wide phenomenon is likely to eclipse the effect of the substantial

range of terrorist ideologies that are contained within a state. This distinction and

ideological competition are discussed in more detail below.

I also restrict my analyses to cases of domestic terrorism. This is important not

only because domestic terror is more prevalent than international terrorism (Rosendorff

and Sandler, 2005), but because analyses that focus solely on international terrorism or

combine the two may lead to incorrect conclusions because they will be biased towards

large states and certain highly capable groups (Abadie, 2006).

In terms of competition, international and domestic groups are likely to face

different pressures. Groups operating domestically have to rely on resources within the

country – meaning that their practices are likely to be heavily driven by their public

support. This means that factors such as competition, as well as the acceptability of

violence and government policy, are likely to show a direct impact on the amount of

terrorism a group perpetrates. An international group, on the other hand, does not face

11

the same pressure. Instead, the international group can circumvent factors at the target

state and go back to its home base to get the needed resources. Competition, then, should

have very little to do with international group behavior; in fact, Siqueira and Sandler

(2006) state that groups with outside sponsors, particularly international groups, are

likely to face little restraints on the type of violence they use.

I create this distinguishing coding rule by noting the attack location and

perpetrating group in the GTD. This is then compared to the available group record in

the START Center’s Terrorist Organizational Profiles (TOPS) database.vii If the location

of the attack in the GTD coincides with a known operating area for the terrorist

organization in the TOPs data, the event is marked as domestic terror.viii The process of

distinguishing international from domestic terror is illustrated in Figure 1.

[Figure 1 Here]

The final dataset contains 31,364 acts of domestic terror perpetrated by 459

terrorist organizations.ix This number is comprised of 429 individual groups plus 30

“franchises”. These attacks are then aggregated by group-state/year to create a dataset of

2,131 group-state/years with terrorist attacks. Using the start and end-dates of terrorist

activity from the TOPS database, I fill in all non-terror years (years the group was active

but not engaged in terrorist attacks) with zeros. The final dataset that results includes

8,217 group-state/years.

Dependent Variable

The dependent variable used in this analysis is a yearly count of the number of

domestic terror attacks committed by each individual terrorist organization. Attacks are

included in the dataset if they are intentional, involve a use of violence or a threatened

12

use of violence, and the actors are sub-national (START, 2009: 4-5). Attacks must also

fulfill two of three additional criteria for inclusion: the act must be directed towards a

political, economic, religious, or social goal, there must be evidence of an intent to

coerce, and the action must be outside of the realm of legitimate military activities

(START, 2009: 4-5). Lastly, to reduce bias that may result from overemphasizing

capable organizations, I follow Maoz (2007) and include failed acts of terrorism.

Independent Variable

To assess the effects of group competition, I use the inverse of the Herfindahl-

Hirschman Index (HHI) as my main independent variable (Herfindahl, 1950; Hirschman,

1945). The HHI was originally conceived as a way to assess the concentration of firms in

a marketplace. Used in this capacity, it has since its inception become a central

component in economic analyses; both the Justice Department and the Federal Reserve

use it to evaluate the market impact of potential mergers (Hannan, 1997; Rhodes, 1993).

One of the advantages of this measure, as opposed to other means of determining

competition, is that it provides a measure that is invariant to the number of competitors.

For example, there may be several firms in a market that create a particular good, yet the

market share of one may create an environment where little competition actually exists.

Similarly, the presence of several terrorist groups may not indicate a truly competitive

environment, especially if one is clearly more capable and more noteworthy than the rest.

The ability of the HHI to account for this potential asymmetry is widely touted as one of

the strengths of the measure (Calkins, 1983).

The HHI is calculated as:

∑i=1

N

S i2

13

Where Si is the market share of firm i and N is the number of firms in the market.

To determine the number of firms in the market, I adopt the concept of “competitive

exclusion” from the organizational ecology literature (Lowery and Gray, 1995). This

states that competition is a function of the resources that one draws from: groups will

only compete with other groups that draw on the same resource (Lowery and Gray,

1995). For Lowery and Gray’s (1995), this meant competition amongst interest groups

occur within similar advocacy markets. Farm advocacy groups are distinct from business

groups and, as such, are only likely to face competition from other farm groups. As a

result, their analysis focuses on “species-guilds” – “sets of organizations representing

related interests” (1995: 9).

I conceptualize the same for terrorist organizations: each state has a certain

number of ideological “markets” in which recruits and resources are sought and

competition is bounded. Communist terrorist organization should face competition from

groups with similar ideological goals, not dissimilar ones. Secessionists have their needs

best served from a nationalist, rather than a religious, organization. Post et al. (2003:

173) notes terrorist recruitment occurring within these lines: “individuals from strictly

religious Islamic backgrounds were more likely to join Islamist groups, while those who

did not have a religious background might join either a secular or religious group.”

Using this insight, I use the RAND Corporation’s End of Terror dataset (Jones

and Libicki, 2008) to classify each state as having four potential ideological “markets” -

nationalist, religious, left-wing, and right-wing. To determine the number of “firms” in

each of these “markets”, I sum for each state/year the number of other groups that exist

within each of the four ideological categories. Market share is then determined by the

14

percent of attacks committed per state/year by each group within each market. This best

corresponds to the insight that members stay in and recruits join the most active

organizations (Crenshaw, 1985; Oots, 1989; Post et al., 2003).

These two values are used to calculate the HHI for each ideological category for

every state-year. The use of the inverse of the HHI follows Taagepara and Shugart

(1989) and allows for a more readily interpretable variable – scores near 1 indicate a

condition of monopoly while higher scores indicate competitive environments.x If the

outbidding argument is correct, high values of this variable will be associated with more

total attacks. If the moderation perspective is correct, low values will be associated with

more attacks.

[Figure 2 Here]

Figure 2 plots the average level of organizational competition for the ten states

with the highest number of domestic terrorist attacks.xi Bars of one unit length indicate a

condition of monopoly while longer bars indicate progressively more competitive

environments. The colors indicate terrorist ideology as determined by the RAND End of

Terror Dataset (Jones and Libicki, 2008). The results from the chart appear to accord

with our understanding of terrorism. Colombia appears to offer an extremely competitive

environment for left-wing groups while France and the UK offer competitive arenas for

nationalist terror organizations. It is also interesting to note that high number of

uncontested categories; most of the top victimized states are characterized by relatively

monopolistic ideological categories. This finding, with the three caveats mentioned

above, contrasts with Bloom’s (2005) outbidding theory.

15

The three exceptions accord well with our understanding of terrorism in those

states. The high score for left-wing groups in Colombia indicate an environment where

the success of organizations such as the Revolutionary Armed Forces of Colombia

(FARC) and the National Liberation Army (ELN) gave rise to smaller organizations such

as the Guevarista Revolutionary Army and the Popular Liberation Army. In France and

the UK, the high level of competition occurs due to groups engaged in struggles for

Basque, Corsican, and Northern Ireland separatism.

Figure 3 plots the average level of competition by regime type over the time

period of the GTD dataset. These are calculated from the Polity IV dataset using the

Polity2 measure (Marshall et al., 2009).xii States with Polity2 scores at 6 and above are

coded as democratic and those below 6 are coded as autocratic. Figure 4 plots

competition by regime type using a trichotomous classification between autocracy,

anocracy, and democracy. The coding rule for democracy and autocracy remain the

same. The middle categorization, anocracy, is coded by Polity2 scores ranging from -5

and 5.

[Figures 3 and 4 Here]

Results indicate that democratic states are more competitive than their autocratic

counterparts (t=3.92, p<.01) and, using the trichotomous classification of regime type,

more competitive than either autocracies or anocracies (F=9.00, p<.01). While the

underrepresentation of terrorist groups in autocracies is a problem for analysis (Drakos

and Gofas, 2006), the difference in the number of groups between the regime types

provides support for Eubank and Weinberg’s (1994, 1998) findings that democracies are

more likely to harbor terrorist groups than autocracies. Further, this echoes Chenoweth’s

16

(2010) research linking the emergence of terrorist organizations in democracies to high

levels of intergroup competition.

Because outbidding theory is a contextual one – depending on the “domestic

politics of the minority group and the state counter-terror strategies and responses to

insurgent violence” (Bloom 2005: 79) – I include a variable assessing the “environment”

that each state provides a terrorist organization. This variable is drawn from Mullins and

Young (forthcoming), who argue that states and societies that legitimize and rationalize

the use of violence may be more likely to be a victim of terrorism. This occurs as a result

of a spillover process – groups model and adopt the state’s use of violence to its own

interactions with others.

This measure is drawn from four distinct indicators – two which capture state

violence against its people, a third that measures citizen violence, and a final measure of

state involvement in war. The first measures whether or not a state used capital

punishment in a given year while the second captures the state’s use of extrajudicial

killings through the Political Terror Scale (Gibney and Dalton, 1996). The third

component assesses citizen violence by using the number of homicides per 100,000

people as drawn from the World Health Organization. The final element simply

measures state participation in an interstate or intrastate war as measured by the

Correlates of War project. Because the four components were found to load on a single

factor - this analysis, like the later analyses in Mullins and Young (forthcoming), will use

a single variable – culture of violence – to conduct the analyses. The temporal range of

this variable reduces the subsequent analyses to the 1970-1997 period.

17

To best assess this relationship, I create an interaction term between the

concentration score and the culture of violence variable. While some information can be

gained from the coefficients on interaction terms, I follow Brambor et al.’s (2006)

suggestion to visually analyze interaction terms. To do this, I use

Boehmke’s (2006) Grinter data utility.xiii

Control Variables

I consider a number of control variables that may have an impact on group

activity. I first consider the state’s ability to exert control over its population and

territory. A weak central authority creates gaps in state control that allow insurgency to

flourish and allows groups to divert resources from organizational security to the conduct

of international terror (Takeyh and Gvosdev, 2002; Lai, 2007). For domestic terrorist

organizations, a weak central government should operate in both of these ways; groups

can operate unencumbered by state repression and they can dedicate more of their

resources to attacks within the state.

I assess state control by using a variable indicating a state’s relative political

capacity (RPC) (Arbetman and Johnson, 2008). This variable was originally used in

Organski and Kugler (1980) to measure the ability of states to wage war independent of

the absolute size of their resource base. At the domestic level, similar measures of

resource extraction have been associated with a decrease in internal violence (Benson and

Kugler, 1998). This variable is roughly calculated as the ratio of the total value of actual

state extractions relative to the potential value of all state extractions (Arbetman and

Johnson, 2008).xiv I anticipate that increased state control will be associated with fewer

attacks.

18

A second control variable I include is the state’s regime type. The impact of

regime type has been thoroughly debated in previous research, resulting in three schools

of thought. The first argues that the liberal characteristics of democracies, such as respect

for civil liberties, freedom of the press, and rights of due process reduce operational costs

for terrorist groups and allow them to recruit new members and plan attacks with minimal

interference from the government (Crenshaw, 1981; Hamilton and Hamilton, 1983;

Schmid, 1992). A second argues that democracies face reduced risks from terrorism

because there are many ways, such as voting, demonstrations, or the formation of

political parties to effect change without the need for violence (Eubank and Weinberg,

1994; Eyerman, 1998). Lastly, the third argues that the components of democracy have

differential effects on terrorism; characteristics like political efficacy and electoral

participation help to reduce terrorism while constraints of executive power increase the

probability of terrorism (Li, 2005). For this, I use regime type data from POLITY IV.xv

I divide this score into three parts and exclude autocracy as the baseline category.

A number of other control variables are used in this analysis. Civil war should

have a positive effect on all three of our dependent variables. Governments engaged in

civil wars should be less able to maintain control over their territory, reducing their

ability to combat terrorism and allowing for groups to devote more resources towards

terrorist violence. At the level of the populace, civil war can sway unaligned moderates

to support terrorist organizations increasing their vitality and longevity (Kalyvas, 2004).

Data for this comes from the UCDP/PRIO Armed Conflict Dataset (Gleditsch et al.,

2001).xvi

19

Larger populations are also likely to have an effect on terrorist group operations.

Large populations provide organizations with a larger base to recruit from, a broader pool

to draw resources from, and a more difficult environment for a state to monitor. All of

these characteristics should contribute to more active organizations. The values for the

population variable are taken from Gleditsch (2002). This variable is logged to reflect

the decreasing benefit to terrorist groups given larger populations.

Lastly, I account for economic development as a final control variable. While

economic discontent has been linked to a variety of political outcomes, research on the

role of economics on terrorism has found little support for a direct effect (Krueger and

Maleckova, 2003; Abadie, 2006; Piazza, 2006). Rather, support and recruitment for

terrorism seem to occur more often amongst the educated and employed than the poor

(Krueger and Maleckova, 2003; Bueno de Mesquita, 2005). At the same time, wealthier

states that are able to provide adequate social welfare policies are less likely to have

popular support for terrorism and to experience terrorism on their soil (Burgoon, 2006).

This is unclear at the level of the group; organizations in low-wealth countries may

produce less terrorism because it is difficult to assemble enough recruits who meet a

group’s qualifications. As the wealth of a state increases, groups may produce more

terrorism. At the highest levels of state wealth, the production of terrorism may taper off

as qualified recruits exist but support for terrorism is low. I account for this by including

the logged term of per capita GDP. Data for this variable comes from Gleditsch (2002).

Because the hypotheses relate to terrorist group activity, as measured by number

of attacks, I use an event count model. Poisson models are not appropriate in this case, as

they assume that the data are independent and homogenous; it is likely that targeting is

20

not independent as groups that have gained experience with it are more likely to use it in

the future (Jackson et al., 2005). Instead, given that the number of total attacks is

dependent and is over-dispersed, I employ a negative binomial model (Long, 1997;

Cameron and Trivedi, 1998). I lagged the independent variables to account for

endogeneity. Lastly, I also include robust standard errors clustered by group to address

any potential problems with heteroskedasticity and serial correlation (Greene, 2002).

RESULTS

[Table 1 Here]

Table 1 presents the results from our test of competition and terrorist activity.

The first model indicates that competition amongst terrorist organizations within their

respective ideological niches has a negative and statistically significant effect on the

number of terrorist attacks that occur on a yearly basis. The substantive effects confirm

this relationship; a one standard deviation increase (.960) in competition yields a nearly

50% decrease in the number of terrorist attacks.xvii Given that the average number of

terrorist attacks in the data is 7.94, groups in environments with competition scores one

standard deviation above the mean (1.69) – an environment with one additional group –

will perpetrate four fewer attacks than it would in a less competitive environment.

Unlike Bloom (2005), this initial finding seems to suggest that moderation, rather than

outbidding, may be the way that terrorist organizations respond to competition.

The controls generally perform as expected. A state’s orientation towards

violence is statistically significant and associated with a greater number of terror attacks,

echoing Mullins and Young (forthcoming). The effect of democracy is also positive and

21

strongly significant. This may indicate that domestic terror is strongly influenced by the

democratic process – issues of representation or legislation may provoke groups to

violence. At the same time, this may also reflect the deep divisions in reporting terrorist

acts between democracies and autocracies (Drakos and Gofas, 2006). GDP per capita is

also shown to be positive and significant. This suggests that groups will be more active

in areas where they can get the best recruits. As expected, relative political capacity is

negative and significant, indicating that states with greater control over their territories

experience less terrorism. This supports research by Fearon and Laitin (2003) and Piazza

(2008) who find that higher levels of state capacity are related to decreased levels of both

insurgency and terrorism, respectively. Lastly, civil war is significant and in the

anticipated direction. This result indicates that, like Kalyvas (2004), we can see domestic

terrorism as a tactic used by combatants within civil war. This may also occur because

civil conflict allows terrorist groups room to operate with less threat of government

repression.

Contrary to our expectations, larger populations are associated with groups

committing less terrorist acts. In fact, population increases of one standard deviation

result in a 45% reduction in group attacks. This result is puzzling; populations accord

larger bases for recruitment, greater target selection, a larger audience to intimidate, and

greater monitoring difficulties for the government. One potential explanation for the

result is that it may be a methodological artifact – population may affect groups of

different ideologies in different ways. I conduct analyses disaggregating group type and

discuss this point in detail below.

22

The second model in Table 1 incorporates the interaction between group

concentration and a state’s acceptance of violence. The interaction effect is positive but

not statistically significant. When analyzed visually using the grinter utility (Boehmke,

2006), as shown in Figure 5, a different interpretation emerges. Here, the effect of

competition on the number of attacks has a slight positive effect as the state the terrorist

group resides in becomes more tolerant of violence - a similar, albeit weaker, finding to

that of Bloom (2005). The controls provide comparable results to that of the first model.

Table 2 disaggregates the analyses into the four ideological categories used by the

Jones and Libicki (2008) data. This may help to address the weak findings presented

above. It also helps to clarify whether the effect of outbidding is different across

ideological types – an element that is not directly addressed by Bloom (2005). One may

anticipate that groups with a religious ideology may be more likely to engage in

outbidding regardless of the acceptability of violence because their spiritual mandate

trumps any desire to gain widespread public support (Juergensmayer, 2003). Nationalist

groups, on the other hand, may be sensitive to outbidding since excessive violence may

negatively affect public support, making the attainment of their goals more difficult

(Sanchez-Cuenca and de la Calle, 2009).

The results indicate that the effects of many of the variables are conditional on

group ideology. This is first echoed in the results for competition; 3 out of the 4 models

demonstrate that competition in the ideological environment leads to fewer attacks. The

lack of a finding in the fourth case - right wing groups may indicate that these types of

groups are less sensitive to issues of citizen perception given that they are, in many

instances, an arm of state policy and a recipient of state support.

23

The conditional effects are made more evident by visually analyzing the

interaction terms in Figure 6. Here, for religious and nationalist terrorist organizations,

the marginal effect of competition on domestic terror attacks is shown to increase as the

acceptability of violence increases.xviii In other words, religious organizations and their

nationalist counterparts respond to competition using outbidding across most states. Left-

wing groups do not appear to engage in outbidding. Rather, for those instances where

the interaction is significant, competition leads to moderation for left-wing organizations.

The interaction for right-wing organizations is not significant, thus I make no claims

about the relationship.

The control variables also indicate that ideology has a considerable effect on their

direction and significance. The state’s culture of violence has a differential effect across

the ideologies; religious groups are less likely to perpetrate attacks in more violence-

acceptant states while both left and right-wing groups are more likely to. For anocracies,

nationalist groups increase their attacks, while left and right-wing groups curtail their

activities. In addition, nationalist and left-wing groups are also active in democracies,

although the results for left-wing organizations fall outside traditional significance levels.

Per capita GDP only has an effect for nationalist groups; nationalist terrorist

organizations such as MEND (Movement for the Emancipation of the Niger Delta) and

GAM (Free Aceh Movement) have used arguments based on the wealth of the central

government to justify their secessionism (Ross, 2004). Relative political capacity

decreases the number of nationalist and left-wing attacks while increasing the number of

right-wing attacks. Given that this measures a government’s ability, a positive

relationship here may denote the inclination of some governments to use right-wing

24

groups as a coercive element of their policy (Mason and Krane, 1989). The civil war

variable is positive and significant for religious and nationalist groups. This may indicate

that these groups have larger goals, such as regime change, than groups of other

ideologies and that a condition of warfare may be the best context to pursue these goals

(). Lastly, population is only significant for left-wing organizations.

CONCLUSION

While theories of outbidding have been central to explanations of terrorist actions

for many years, few attempts have been made to subject the theories to empirical testing.

This analysis, focusing on group level data, finds a positive relationship between

competition and increases in terrorist violence, conditional on a state’s acceptance of

violence. As such, this finding is consistent with those of Bloom (2005). These results

become more pronounced when the analysis takes group’s ideologies into account.

Religious terrorist groups show the greatest proclivity to outbid followed by nationalist

organizations. The positive finding for religious organizations affirms Juergensmeyer’s

(2003: 221) observation that “there is no need…to contend with society’s laws and

limitations when one is obeying a higher power”. Left-wing organizations respond by

reducing their violence – precisely responding to society’s limitations.

In addition, this analysis utilized the organizational ecology literature to better

operationalize the concept of competition. Previous discussions of outbidding and

competition treated competition as boundless – resources and recruits ran towards the

most capable organization. Similarly, recruits were drawn to any group, regardless of

their ideology or history. Interviews and assessments of terrorist operatives have shown

that this is not the case, recruits and resources are very specific about their destination

25

(Post et al., 2002). As such, terrorist organizations can be theorized to operate via the

“competitive exclusion” principle – competition only occurs between organizations of a

similar type. This then allows us to create measures that take this type of sectoral

competition into account.

This research also points to the value of considering using the group as the unit of

analysis. Work by Asal and Rethemeyer (2008) demonstrate this utility, showing that

some organizational attributes have important impacts on organizational lethality while

others do not. Here, this choice shows various levels of competition, and thus very

disparate levels of violence, can exist within a state. Aggregating the data to use larger

units of analysis, like the state, threatens to obscure this valuable heterogeneity. Such a

focus, and the utilization of concepts from different fields, may provide more clarity to

this complex and menacing phenomena.

26

Figure 1: Coding Rule & Screening Mechanism

Case Group Date State

1 Farabundo Marti National Liberation Front (FMLN)

1/14/1980 El Salvador

2 FMLN 1/24/1980 El Salvador

3 FMLN 2/14/1980 Guatemala

4 FMLN 9/15/1980 El Salvador

5 FMLN 2/4/1981 El Salvador

6 FMLN 11/12/1981 El Salvador

7 FMLN 4/26/1982 El Salvador

8 FMLN 10/18/1982 El Salvador

9 FMLN 2/3/1983 El Salvador

10 FMLN 8/18/1983 United States

27

Selected FMLN Attacks 1980-1983

Step 2. These states are not defined as regular operating areas for the FMLN and, thus, are not considered domestic.

Step 3. As a result, Cases 3 and 10 are not coded as cases of domestic terrorism.

Step 1. According to TOPs, the FMLN was known to operate solely in El Salvador.

Figure 2: Average Level of Concentration for Top Ten Target States of Domestic Terrorism (1970-1997)

0.5

11.

52

2.5

Chile

Colombia

El Salv

ador

Franc

ePeru

Philipp

ines

Spain

Sri Lan

ka

Turke

y UK

Religious NationalistLeft-Wing Right-Wing

28

Figure 3: Average Level of Terrorist Organization Concentration (Autocracies and Democracies) (1970-1997)

1.2

1.4

1.6

1.8

22.

2C

ompe

titio

n In

dex

1970 1980 1990 2000Year

Democracy Autocracy

29

Figure 4: Average Level of Terrorist Organization Concentration (Autocracies, Democracies, and Anocracies) (1970-1997)

11.

52

2.5

3C

ompe

titio

n In

dex

1970 1980 1990 2000Year

Democracy AnocracyAutocracy

30

Figure 5: Effect of Competition Conditioned on Acceptability of Violence

0.1

.2.3

.4K

erne

l Den

sity

Est

imat

e of

lagf

acto

r

-1.5

-1-.5

0

Mar

gina

l Effe

ct o

f Com

petit

ion

on N

umbe

r of D

omes

tic T

erro

r Atta

cks

-1 0 1 2 3Acceptability of Violence

31

Figure 6: Effect of Competition Conditioned on Acceptability of Violence by Group Type

32

Table 1: The Effect of Competition on Domestic Terrorist Attacks Model 1 %Δ Model 2 %Δ

Competition -.714***(.122)

-49.6 -.770***(.166)

-52.3

Culture of Violence .752***(.151)

124.2 .656**(.294)

102.1

Concentration * Culture .055(.103)

Anocracy .019(.391)

-.002(.385)

Democracy 1.25***(.419) 247.9 1.23***

(.422) 240.7

Logged GDP per capita .000***(.000)

145.4 .000***(.000)

148.8

RPC -.506*(.261)

-23.6 -.503*(.261)

-23.5

Civil War .750**(.310)

111.8 .764**(.310)

114.7

Logged Total Population -.379***(.113)

-45.3 -.383***(.113)

-45.7

Constant -4.63(1.29)

-4.74(1.31)

N 3111 3111Log Likelihood -5900.73 -5900.12Wald χ2 95.38*** 121.11***Note: Robust Standard Errors clustered on group in parentheses. * significant at 0.10, ** significant at 0.05, *** significant at 0.01 (two tailed). All independent variables lagged at t-1. Percentage change calculated as change in 1 standard deviation for continuous variables and unit change for dichotomous variables.

33

Table 2: The Effect of Competition on Total Terrorist Attacks by IdeologyReligious %Δ Nationalist %Δ Left Wing %Δ Right Wing %Δ

Competition -1.69***(.608)

-76.4 -.824***(.164)

-60.5 -.526**(.241)

-32.0 -.321(.330)

Culture of Violence

-2.03***(.633)

-86.7 .215(.306)

.911***(.336)

177.2 1.27**(.616)

283.2

Con * Culture .994***(.367)

591.8 .110(.092)

-.063(.137)

-.527(.458)

Anocracy -.284(.662)

2.00***(.374)

637.1 -.840**(.403)

-56.8 -2.92***(.562)

-94.6

Democracy -.018(.673)

1.84***(.330)

529.4 .949*(.542)

158.4 .288(.709)

Logged GDP per capita

-.000(.000)

.000***(.000)

294.3 .000(.000)

-.000(.000)

RPC -.226(.979)

-.847**(.351)

-39.5 -.857**(.387)

-29.8 .495**(.228)

28.5

Civil War 1.62**(.801)

403.4 .742**(.290)

109.9 .658(.474)

.903(.583)

Logged Total Population

-.374(.266)

.161(.111)

-.647***(.218)

-60.5 -.264(.385)

Constant 6.78(2.57)

-1.91(1.41)

8.28(2.21)

2.67(3.37)

N 422 1371 1149 169Log Likelihood -498.77 -2328.64 -2621.95 -209.31Wald χ2 25.46*** 114.81*** 107.10*** 453.13***

Note: Robust Standard Errors clustered on state in parentheses. * significant at 0.10, ** significant at 0.05, *** significant at 0.01 (two tailed). All independent variables (other than ideology) lagged at t-1. Percentage change calculated as change in 1 standard deviation for continuous variables and unit change for dichotomous variables.

34

References

Abadie, Alberto. 2006. Poverty, Political Freedom, and the Roots of Terrorism. American Economic Review 96(2): 50-56.

Arbetman, Marina and Kristin Johnson (2008) Relative Political Capacity: Empirical and Theoretical Underpinnings. Paper presented at the 49th Annual Convention of the International Studies Association, San Francisco, CA, 26–29 March.

Asal, Victor and R. Karl Rethemeyer. 2008. The Nature of the Beast: Organizational Structure and the Lethality of Terrorist Attacks. Journal of Politics 70(2): 437-449.

Benson, Michelle and Jacek Kugler. 1998. Power Parity, Democracy, and the Severity of Internal Violence. Journal of Conflict Resolution 42(2): 196-209.

Bloom, Mia. 2005. Dying to Kill: The Allure of Suicide Terror. New York: Columbia University Press.

Boehmke, Frederick. 2006. “Grinter: A Stata Utility for Graphing the Marginal Effect of

an Interacted Variable in Regression Models.” http://myweb.uiowa.edu/fboehmke/methods.html (January 23, 2010)

Brambor, Thomas, William Clark, and Matt Golder. 2006. Understanding Interaction

Models: Improving Empirical Analyses. Political Analysis 14(1): 63-82.Brym, Robert and Bader Araj. 2008. Palestinian Suicide Bombing Revisited: A Critique

of the Outbidding Thesis. Political Science Quarterly 123(3): 485-500 Bueno de Mesquita, Ethan. 2005. The Quality of Terror. American Journal of Political

Science 49(3): 515-530.Bueno de Mesquita, Ethan. 2007. Politics and the Suboptimal Provision of Counterterror.

International Organization 61(1): 9-36.Burgoon, Brian. 2006. On Welfare and Terror: Social Welfare Policies and Political-

Economic Roots of Terrorism. Journal of Conflict Resolution 50(2): 176-203.Calkins, Stephen. 1983. The New Merger Guidelines and the Herfindahl-Hirschman

Index. California Law Review 71(2): 402-429.Cameron, Colin and Pravin Trivedi. 1998. Regression Analysis of Count Data. Oxford:

Oxford University Press. Chenoweth, Erica. 2010. Democratic Competition and Terrorist Activity. Journal of

Politics 72(1): 16-30.Coser, Lewis. 1956. The Functions of Social Conflict. Glencoe: Free Press. Crenshaw, Martha. 1981. The Causes of Terrorism. Comparative Politics 13(4): 379-

399. Crenshaw, Martha. 1985. An Organizational Approach to the Analysis of Political

Terrorism. Orbis 29(3): 465-489.Cronin, Audrey Kurth. 2006. How al-Qaida Ends: The Decline and Demise of Terrorist

Groups. International Security 31(1): 7-48.Dingley, James. 2001. The Bombing of Omagh, 15 August 1998: The Bombers,

35

Their Tactics, Strategy, and Purpose Behind the Incident. Studies in Conflict and Terrorism 24: 451-465.

Drakos, Konstantinos and Andreas Gofas. 2006. The Devil You Know but Are Afraid to Face: Underreporting Bias and its Distorting Effects on the Study of Terrorism. Journal of Conflict Resolution 50(5): 714-735.

Eubank, William and Leonard Weinberg. 1994. Does Democracy Encourage Terrorism? Terrorism and Political Violence 6(4): 417-443.

Eubank, William and Leonard Weinberg. 1998. Terrorism and Democracy: What Recent

Events Disclose. Terrorism and Political Violence 10(1): 108-118.Eyerman, Joe. 1998. Terrorism and Democratic States: Soft Targets or Accessible

Systems. International Interactions 24(2): 151-170. Fearon, James and David Laitin. 2003. Ethnicity, Insurgency, and Civil War. American

Political Science Review 97(1): 75-90.Gibney, Mark, and Matthew Dalton. 1996. The Political Terror Scale. Policy Studies

and Developing Nations 4: 73-84. Gleditsch, Nils Petter, Peter Wallenstein, Mikael Eriksson, Margareta Sollenberg, and

Håvard Strand. 2001. Armed Conflict, 1946-2001: A New Dataset. Journal of Peace Research 39(5): 615-637.

Gleditsch, Kristian. 2002. Expanded Trade and GDP Data. Journal of Conflict Resolution 46(5): 12-24.

Greene, William. 2002. Econometric Analysis. 5th ed. Upper Saddle River: Prentice

Hall.Hamilton, Lawrence and James Hamilton. 1983. Dynamics of Terrorism. International

Studies Quarterly 27(1): 39-54.Hannan, Timothy. 1997. Market Share Inequality, the Number of Competitors and the

HHI: An Examination of Bank Pricing. Review of Industrial Organization 12(): 23-35.

Herfindahl, Orris. 1950. Concentration in the Steel Industry. Ph.D. Diss., Columbia University.

Hirschman, Albert. 1945. National Power and the Structure of Foreign Trade. Berkeley:

University of California Press.Hoffman, Bruce. 1998. Inside Terrorism. New York: Columbia University Press.Jackson, Brian, John Baker, Kim Cragin, John Parachini, Horacio Trujillo, and Peter

Chalk. 2005. Aptitude for Destruction: Organizational Learning in Terrorist Groups and its Implications for Combating Terrorism. Santa Monica: RAND Corporation

Jacobson, Philip. 2001. “Home-Grown Martyrs of the West Bank Reap Deadly Harvest.”

Sunday Telegraph, 19 August. Jones, Seth and Martin Libicki. 2008. How Terrorist Groups End: Lessons for

Countering al Qa’ida. Santa Monica: RAND Corporation Juergensmeyer, Mark. 2003. Terror in the Mind of God. Berkeley: University of

36

California Press.Kalyvas, Stathis. 2004. The Paradox of Terrorism in Civil War. Journal of Ethics 8(1):

97-138. Kellen, Konrad. 1979. Terrorists-What are they Like? How Some Terrorists Describe

Their World and Actions. Santa Monica: RAND Corporation.Krueger, Alan and Jitka Maleckova. 2003. Education, Poverty, and Terrorism: Is There

a Causal Connection? Journal of Economic Perspectives 17(4): 119-144.Kydd, Andrew and Barbara Walter. 2006. The Strategies of Terrorism. International

Security 31(1): 49-80.LaFree, Gary and Laura Dugan. 2007. GTD1 (1970-1997) Global Terrorism Database

Codebook. College Park: START. Lai, Brian. 2007. ‘Draining the Swamp’: An Empirical Examination of the Production

of International Terrorism, 1968-1998. Conflict Management and Peace Science 24(4): 297-310.

Li, Quan. 2005. Does Democracy Promote or Reduce Transnational Terrorism Incidents?

Journal of Conflict Resolution 49(2): 278-297. Long, J. Scott. 1997. Regression Models for Categorical and Limited Dependent

Variables. Thousand Oaks, CA: Sage Publications.Long, J. Scott and Jeremy Freese. 2005. Regression Models for Categorical Outcomes

Using Stata. Second Edition. College Station, TX: Stata Press.Lowery, David and Virginia Gray. 1995. The Population Ecology of Gucci Gulch,

or the Natural Regulation of Interest Group Numbers in the American States. American Journal of Political Science 39(1): 1-29.

Maoz, Zeev. 2007. Evaluating Israel’s Strategy of Low-Intensity Warfare, 1949-2006. Security Studies 16(3): 319-349.

Marshall, Monty, Ted Gurr, and Keith Jaggers. 2009. Polity IV Project: Political Regime Characteristics and Transitions, 1800-2009. Center for Systemic Peace.

Moloney, Ed. 2010. Voices from the Grave: Two Men’s War in Ireland. New York: PublicAffairs.

Mullins, Christopher and Joseph Young. Forthcoming. Cultures of Violence and Acts of Terror: Applying a Legitimation-Habituation Model to Terrorism. Crime & Delinquency

Oots, Kent. 1989. Organizational Perspectives on the Formation and Disintegration of Terrorist Groups. Studies in Conflict and Terrorism 12(3): 139-152.

Organski, A F K. and Jacek Kugler. 1980. The War Ledger. Chicago: University of Chicago Press.

Piazza, James. 2006. Rooted in Poverty?: Terrorism, Poor Economic Development, and Social Cleavages. Terrorism and Political Violence 18(1): 159-177.

Piazza, James. 2008. Incubators of Terror: Do Failed and Failing States Promote Transnational Terrorism? International Studies Quarterly 52(3): 469-488.

Post, Jerrold, Ehud Sprinzak, and Laurita Denny. 2003. The Terrorists in Their Own Words: Interviews with 35 Incarcerated Middle Eastern Terrorists. Terrorism and Political Violence 15(1): 171-184.

37

Rhoades, Stephen. 1993. The Herfindahl-Hirschman Index. Federal Reserve Bulletin 79: 188-189.

Rosendorff, B. Peter and Todd Sandler. 2005. The Political Economy of Transnational

Terrorism. Journal of Conflict Resolution 49(2): 171-182. Ross, Jeffrey and Ted Robert Gurr. 1989. Why Terrorism Subsides: A Comparative

Study of Canada and the United States. Comparative Politics 21(4): 405-426.Ross, Michael. 2004. How Do Natural Resources Influence Civil War? Evidence from

Thirteen Cases. International Organization 58(1): 35-67.Sanchez-Cuenca,Ignacio and Luis de la Calle. 2009. Domestic Terrorism: The Hidden

Side of Political Violence. Annual Review of Political Science 12: 31-49Sandler, Todd, John Tschirhart, and Jon Cauley. 1983. A Theoretical Analysis of

Transnational Terrorism. American Political Science Review 77(1): 36-54.Schmid, Alex. 1992. Terrorism and Democracy. Terrorism and Political Violence 4(4):

14-25.Schmid, Alex and Albert Jongman. 1988. Political Terrorism: A Research Guide to the

Concepts, Theories, Databases, and Literature. Amsterdam: North Holland. Silke, Andrew. 2004. An Introduction to Research on Terrorism. In Research on

Terrorism: Trends, Achievements, and Failures, ed. Andrew Silke. London: Frank Cass.

Siquiera, Kevin and Todd Sandler. 2006. Terrorists versus the Government: Strategic

Interaction, Support, and Sponsorship. Journal of Conflict Resolution 50(6): 878-898.

Taagepera, Rein and Matthew Shugart. 1989. Seats and Votes. New Haven: Yale University Press.

Takeyh, Ray and Nikolas Gvosdev. 2002. Do Terrorist Networks Need a Home? The Washington Quarterly 25(3): 97-108.

Waugh, William. 1983. The Values in Violence: Organizational and Political Objectives of Terrorist Groups. Conflict Quarterly 3(4): 5-19

Wolf, John. 1978. Organization and Management Practices of Urban Terrorist Groups. Terrorism: An International Journal 1(2): 169-186.

38

39

i This can also occur as a response to competition within the organization. Moderates may radicalize to prevent radical members from defecting to form a competing organization. ii An example of the consequences of ignoring radicalization is the fracturing of the Popular Front for the Liberation of Palestine in the early 1980s when it refused to consider participating in acts of violence such as hijacking or guerrilla warfare (Crenshaw, 1985: 85). iii See Brym and Araj (2008) for a reexamination of Bloom’s (2005) data and a critique of the outbidding hypothesis. iv It took fourteen months for suicide terrorism to spread from fundamentalist organizations like Palestinian Islamic Jihad (PIJ) and Hamas to secular organizations like the Al-Aqsa Martyr’s Brigade and the Popular Front for the Liberation of Palestine (PFLP) (Brym and Araj, 2008). v Using the online component of the GTD (www.start.umd.edu/gtd/), I gathered the total number of events for the two terrorist organizations. I compared the total number of attacks for each group to the number of terrorist attacks that occurred during each occasion of “cooperation/coordination.” The PKK participated in 994 total incidents in Turkey, 367 of which occurred after the “cooperation/coordination” phase began in 1993. Turkish Hezbollah participated in 4 incidents, all of which occurred after 1993. Altogether, 371 attacks out of 998 (37%) occurred during the cooperative phase. For Hamas and Fatah the “cooperation/coordination” phase is the duration of the Second Intifada (2000-2006). Hamas perpetrated 287 total attacks, 150 of which occurred during the Intifada. Fatah engaged in 52 total attacks, 10 of which occurred during the Intifada. In this case, 160 out of 339 (47%) attacks occurred during the Intifada. vi During the transfer of the files from their original owner to START, the files for 1993 were lost. Analyses using this data are all missing in this time period. vii TOPs is located at http://www.start.umd.edu/start/data/tops/. viii This also holds for terrorist organizations operating in more than one country given that the additional countries are recognized by the data. ix Out of the total number of cases - 81,800 – 32,112 were excluded because they were listed as “unknown” or “other”. From the remainder, 11,882 cases were dropped because they included unorganized collectives such as “a deranged patient”, “student protestors”, or “pirates” as well as groups that are not accounted for in TOPs. A further 2,155 were dropped – 836 because they were doubtful instances of terrorism and 1,319 were dropped because they represented excluded target types. In all, this results in 35,651 total cases; 31,364 domestic and 4,287 international. x This measure is called the “effective number of political parties” and provides an indication of “the number of hypothetical equal-sized parties that would have the same effect on fractionalization of the party system as have the actual parties of varying sizes” (Taagepara and Shugart, 1989: 79). xi Scores for all the states (N = 82) in the analysis are available upon request. xii I use the Polity2 measure from the dataset; this provides a twenty point scale created by subtracting the Autocracy score from the Democracy score while accounting for the effects of transitions, interregnums, and interruptions. “Transitions” (-88) are interpolated across the span of the transition. “Interregnums” (-77) are given a neutral score of 0 and “interruptions” (-66) are coded as missing (Marshall et al., 2009). xiii Grinter works by plotting the marginal effect of the primary independent variable on the conditional variable while holding others constant. xiv For a more detailed discussion of the construction of the variable, refer to Arbetman and Johnson (2008). xv Once again, I use the Polity2 variable. Substitution with the original Polity variable does not appreciably change the results. xvi This dataset defines armed conflict as, “a contested incompatibility that concerns government and/or territory where the use of armed force between two parties, of which at least one is the government of a state, results in at least 25 battle-related

deaths” (UCDP/PRIO Armed Conflict Dataset, 2009: 1). Four types of conflict are coded in the data: extrasystemic, interstate, internal conflict, and internationalized internal conflict. For the purposes of this study, I restricted the analysis to the last two forms of conflict. xvii Substantive effects were calculated using the listcoef function in STATA (Long and Freese, 2005)xviii In the case of religious organizations, this effect is significant only when the upper and lower bounds of the confidence interval exclude zero.